A new journal paper co-authored by Pengshun Li, Ziqi Wang, Bingyu Zhao, Tracy Becker, and Professor Kenichi Soga has launched a novel “simulation‑free” surrogate model to assess bridge criticality within traffic networks under earthquake conditions.
Published in Transportation Research Part D: Transport and Environment, the study addresses a key engineering challenge: optimizing seismic resilience in bridge networks like those in the San Francisco Bay Area without the time‑consuming burden of full traffic simulations. The authors employ a Markov random walk model paired with a random forest surrogate to estimate network performance, along with a combined ranking method—leveraging One‑at‑a‑time sensitivity, Sobol’ indices, and Gini importance—to identify the most critical bridges for retrofit.
Key highlights include:
- 98 % reduction in computation time compared to traditional simulation-based approaches, enabling rapid resilience assessments.
- Strong predictive performance with minimal training data, minimizing data collection effort.
- Improved bridge prioritization, outperforming existing methods when evaluating network performance improvement potential.
This workflow empowers infrastructure planners and owners to quickly and accurately identify key bridge elements that, when strengthened, yield the greatest benefit for network resilience in earthquake scenarios. The study is particularly timely for regional infrastructure planning, offering a scalable decision-support tool with proven real-world applicability.